Pavel's diary

Get ActiveCard USB reader working on GNU/Linux (pcsc's CCID driver is not packed inside bundle in SL10).

Added support for Persian (Farsi) to my build system.

Read some documentation and specs about AltiVec/Velocity Engine in G4/G5 processors. Hmm, the results of running openssl speed on low end-Mac mini (building OpenOffice.org at the same time) and my notebook are comparable (Mini is a bit faster even under load). I had to upgrade. My notebook.

Found out that Mac OS X 10.4.3 has some problems when one service/program tries to connect to firewall-ed remote port. The whole UI has been frozen for some time and I had to Force Quit it (the rainbow baloon is hardware-implemented or what, it was still rotating ;-). I tried to connect to wrong VNC display number on my iMac, my current build system (which is even slower than new Mini...). I hope to finish my first (only manual for now) build on Mini tonight (hoping the disk won't be full before the end).

Heiner solved my Solaris/SPARC issue #i54319# so I can remove my hack to skip checkdll segmentation faults and report/ignore them. -----

Today, I started building on Tiger with gcc-3.3. The first problem was wrong integration for stlport fix #i49044#. It simply is not present in current m142. -----

I was able to use two different USB keyboards (DELL's RT7D50 and Chicony's KU-2971) tonight to bring my Mac Mini to life, so my Genius SlimStar saga is now finished for me. I will use it as a test for Apple developers (how fast they are in fixing such issues ;-)

Enjoyed fast user switching, slow disk, Expose, Dashboard, 200+ MB updates, 1680x1050 resolution, noisy DVD, etc. ;-)

Thanks to Fipa for helping me to get up to speed. But it still needs some tweaking to use MacOS X to be as productive with it as with GNU/Linux desktop. -----

OpenOffice.org RC1 is ready so I have started all builds (Build-2 of SRC680_m142).

Status of GSI files errors:

     10 GSI_da.sdf.err
     11 GSI_km.sdf.err
     12 GSI_mk.sdf.err

The number is the number of wrong lines in GSI files. Please verify respective err files (already mirrored) so the next RC (should there be any) is without errors! -----

Hmm, strange things happen. I'm able to use my wife's Genius SlimStar keyboard in Mac Mini's Open Firmware. I was even able to set my own oem-banner...

So it works in Open Firmware and it does not in the operating system. Thus the kernel is guilty. Is there any known fix or workaround for this problem or do I really have to wait for Monday? ;-)

Update: I have reported this issue as ID# 4355177 in Apple Bug Reporter (even slower than OOo's IZ BTW! I thought it is impossible to achieve this goal...). -----

To get some knowledge of Mac OS X, I now have Apple's Mac Mini at home. I have connected the keyboard and guess what, it is not supported (another link)! ;-)

Yes, I'm lucky enough that my wife's USB keyboard is Genius SlimStar!

First feelings from having my first Mac:

possitive: perfect packing of the packet, impressive design of the box, CDs and docs package, blinking of the LED while in sleep mode, impressive fast wake up from sleep mode.

negative: press a key to identify your keyboard.

Do you know the feeling that you want to play with your new toy and you can't ;-) It is even worse than to not be able to work on your favourite open-source project... -----

What is better: lists with Reply-To mangling so people respond back to the list by default or no lists or no mail at all?

Yes, OpenOffice.org project is again without mail (#i58303#). And as always, it is right before preparations for RC1. This time, it is again more than 24 hours - this time, 31 hours right now! Unbelievable, frustrating, poisonous and disgusting. Give me another dictionary, please!

The worst thing is that operators started to work on it immediately after two hours and it still doesn't work 29 hours later... -----

Fabrice Bellard implemented SMP support into QEMU. Please test it carefully (it was tested on GNU/Linux only - 2.4 and 2.6)!

Another interesting news - Paul Brook announced initial ARM system emulation. -----

Apple's documentation is impressive. I have never seen an operating system (unfortunately non-free, able to run on proprietary hardware only and with strange future) with such systematic, on-line, free (as in free beer) documentation... During this weekend, I have read Mac OS X Technology Overview. -----

I spent a lot of time today to get some knowledge about universal binaries, a concept from Apple, that helps them to keep one single binary able to run on PowerPC based Macintosh computers and also on their new Intel based machines.

Thinking about it deeper. Interesting problems to solve ;-) How would you solve the problem defined as "Prepare ABI interface for two different architectures". Where different means: different instruction sets and instructions aligning, endianity, argument passing on stack or in registers, many and few registers, different results of divide-by-zero etc.

Funny citation from Universal Binary Programming Guidelines, 2nd edition (taken out of context):

Here are the most typical behavior problems you'll observe when your application runs natively on an Intel-based Macintosh computer: 1. The application crashes. -----

This morning, I attended Sun Galaxy roadshow. Interesting presentations about new Opteron based Sun servers (X2100, X4[12]00) and also some details about Sun Ultra 20 workstation and VMware software. All presentations were well prepared and not pure marketing ;-) On the other hand, the pure technical details (lahf and sahf AMD64 platform instructions on EM64T, see this article for more details) were used as a pure marketing tool ;-) But all in all very interesting morning.

I met Pavel Lisý again after a year approx.

After thinking about current position of VMware, Xen and other virtualizers, I wonder when x86 hardware producers implement the possibility to create virtual machines directly to the hardware/BIOS. Imagine starting a completely new PC server you just bought:

No virtual machine defined. Please create one. Choose one of the following:
     Single machine         Wizard

Is it a dream only? No, this already works. Unfortunately not for x86/AMD64 :-( Why? When? -----

SRC680_m140 is ready. POT files uploaded. My patch tree reduced significantly - many patches/child workspaces integrated.

My builds stopped in module svtools. The same for Caolan and he was also faster when filing an issue ;-) See #i57954# for more details (including not yet delivered files is not possible. It only works for Hamburg RE as a random coincidence). Ause promised to fix this on the master tomorrow.

Second issue (Missplaced close tag) is connected with commented region in one .xcu file. Ivo pointed us to issue #i57889#.

Ivo also notified us that child workspace c06 introduced some new German strings not encoded in UTF8, thus they won't be displayed correctly in the user interface.

BTW: new strings. I updated my PO tree to new POT files and realized that there are at least 9 new strings. It is OK for community translators using community build system (because they can update their GSI to the very last point), but already merged GSI files will be outdated immediately after the released. We have to changed either the cws integration/rejecting rules or the mechanisms for including GSI files. I'd like to hear your opinion. -----

I never thought I'll spend so much time on debugging an error like this one. See #i54319# for more details and help us, please ;-)

Other work today was mostly administrative - meetings, testing, meetings, reshuffling build servers, reading some facts to be prepared for Sun Galaxy presentation tomorrow (more details tomorrow) etc. -----

The milestone SRC680_m139 is ready. The only problem I found during building on GNU/Linux is #i57713#. Build in module instsetoo_native is trying to build SDK installer, but module sdk_oo is not a prerequisite of any other module, thus the file sdk_oo.zip is missing:

ERROR: File not found: sdk_oo.zip

Workaround is simply, just build module sdk_oo and then module instsetoo_native again.

Wrong lines in GSI files:

   323 GSI_da.sdf.err
    23 GSI_el.sdf.err
    11 GSI_km.sdf.err
    84 GSI_mk.sdf.err

All error files uploaded.

I have also extended my script for downloading GSI files. It is now checking the correctness of used ISO code. Several issues revealed. Both Serbian files use ISO code sr and Nepali GSI contained only en-US lines for several strings. -----

I have extended my build system to produce better error output for translators. Previous GSI.sdf.err files contained only lines with bugs, new err files will also contain description of the problem.

For example, see the err file GSI_da.sdf.err. Its first line contains this:

Error: Translation, Line 35936, UniqueID
helpcontent2/source\text\sdraw\guide\text_enter.xhp/help/par_id3155266/49//:
Missing quotes ( \" ) in Tag: .... height=\"5.64mm"\> ...

It is pretty clear that the translator exchanged backslash and quote at the end of height attribute.

I'll upload GSI err files with every build as I did before. -----

... for 9 hours right now (#i57472#).

Update (Mon Nov 7 20:34:44 CET 2005): 10 hours right now.

Update (Mon Nov 7 22:23:40 CET 2005): 12 hours without mail.

Update (Tue Nov 8 06:37:08 CET 2005): 20 hours without mail.

Update (Tue Nov 8 08:10:21 CET 2005): ~22 hours without mail.

Update (Tue Nov 8 09:26:05 CET 2005): ~23 hours without mail. Only one hour missing to be one complete day without mails.

Update (Tue Nov 8 10:47:37 CET 2005): we reached one day milestone. Unbelievable. -----

My current build is SRC680_m138. Builds for GNU/Linux and Windows uploaded. Several issues on MacOS X (agg/canvas related).

Several GSI files are not properly updated:

pavel@paveljanik:/tmp/GSI> grep -l OOO_LICENSE *
GSI_bg.sdf
GSI_el.sdf
GSI_en-GB.sdf
GSI_fi.sdf
GSI_ga.sdf
GSI_lo.sdf
GSI_ne.sdf
GSI_sh-YU.sdf
GSI_sr-CS.sdf
pavel@paveljanik:/tmp/GSI>

The tag OOO_LICENSE is no longer in English original string, so it has to be removed from your translations. Please remove it from your GSI files. Otherwise you'll end up with English About window (because I'm preprocessing all GSI files to remove such lines ;-). -----

Kendy is doing a lot of work on 64bit porting and we (me and Dan) finally found why my dual Opteron system segmentation faults in every second grep call ;-) It was probably broken power supply, so I'm now working on dual Opteron without one of the two Opterons ;-) But it doesn't matter it is almost as fast as with two of them...

I started to catch up Kendy's porting efforts and reported/fixed Mozilla's 64bit issue #i57409# and reported/cwsed my fault (-fno-use-cxa-atexit, #i57410#). -----

The milestone SRC680_m138 is ready. Do not forget to check if the new module agg is checked out.

If you build using gcc-3.3 or on FreeBSD or MacOS X, have a look at #i57395#. -----

I had a funny day. It started with the visit of Apple Center with Filip Molčan, my wife and my son ;-) My wife likes Mac mini and iPod nano ;-) The design of Apple Cinema displays is impressive.

Meeting with Filip and Martin about local marketing of OpenOffice.org. CWS stuff (pj39), updating GSI files, Build-3 of SRC680_m137 started.

General meeting of Rudna.net. -----